static inline bool avxcontains(hashset_t * set, uint64_t target) {
__m256i vtarget = _mm256_set1_epi64x(target);
__m256i vlocation = _mm256_and_si256(avxhash(vtarget, set->vmultiplier),set->sizemask);
__m256i svalue = _mm256_i64gather_epi64((const long long int *) set->data,vlocation,8);
__m256i eq = _mm256_cmpeq_epi64(vtarget,svalue);
return _mm256_testz_si256(eq,eq) == 0;
}
int main() {
hashset_t H;
for(int k = 0; k < K; ++k) H.multiplier[k] = getrand64();
H.vmultiplier= _mm256_loadu_si256((__m256i const * )H.multiplier);
uint64_t howmany = 100;
uint64_t * keys = malloc(sizeof(uint64_t) * howmany);
for(uint64_t k = 0; k < howmany; ++k) keys[k] = getrand64();
for(H.size = 1024; H.size < (UINT64_C(1) << 32) ; H.size *=2) {
H.sizemask = _mm256_set1_epi64x(H.size-1);
printf("alloc size = %f MB \n", H.size * sizeof(uint64_t) / (1024 * 1024.0));
H.data = calloc(H.size , sizeof(uint64_t));
for(int j = 0; j < howmany; j += 2) H.data[hash(H.multiplier[0],j) & (H.size - 1)] = j;
int answer = expected(&H,howmany,keys);
RDTSC_BEST(checkthemall(&H,howmany,keys), answer, cache_flush(&H,howmany,keys), 50,howmany);
RDTSC_BEST(avxcheckthemall(&H,howmany,keys), answer, cache_flush(&H,howmany,keys), 50,howmany);
free(H.data);
}
free(keys);
}
/*!
* \brief Fill a packed vector by replicating a value
*/
ETL_STATIC_INLINE(avx_simd_long) set(int64_t value) {
return _mm256_set1_epi64x(value);
}
static FORCE_INLINE void FlowInter_8px_AVX2(
int w, PixelType *pdst,
const PixelType *prefB, const PixelType *prefF,
const int16_t *VXFullB, const int16_t *VXFullF,
const int16_t *VYFullB, const int16_t *VYFullF,
const uint8_t *MaskB, const uint8_t *MaskF,
int nPelLog,
const __m256i &dwords_time256, const __m256i &dwords_256_time256,
const __m256i &dwords_ref_pitch, const __m256i &dwords_hoffsets) {
__m256i dwords_w = _mm256_add_epi32(_mm256_set1_epi32(w << nPelLog), dwords_hoffsets);
__m256i dstF = lookup_AVX2(VXFullF, VYFullF, prefF, w, dwords_time256, dwords_ref_pitch, dwords_w);
__m256i dstB = lookup_AVX2(VXFullB, VYFullB, prefB, w, dwords_256_time256, dwords_ref_pitch, dwords_w);
__m256i dstF0 = _mm256_i32gather_epi32((const int *)prefF, dwords_w, sizeof(PixelType));
__m256i dstB0 = _mm256_i32gather_epi32((const int *)prefB, dwords_w, sizeof(PixelType));
dstF0 = _mm256_and_si256(dstF0, _mm256_set1_epi32((1 << (sizeof(PixelType) * 8)) - 1));
dstB0 = _mm256_and_si256(dstB0, _mm256_set1_epi32((1 << (sizeof(PixelType) * 8)) - 1));
__m256i maskf = _mm256_cvtepu8_epi32(_mm_loadl_epi64((const __m128i *)&MaskF[w]));
__m256i maskb = _mm256_cvtepu8_epi32(_mm_loadl_epi64((const __m128i *)&MaskB[w]));
const __m256i dwords_255 = _mm256_set1_epi32(255);
__m256i maskf_inv = _mm256_sub_epi32(dwords_255, maskf);
__m256i maskb_inv = _mm256_sub_epi32(dwords_255, maskb);
__m256i dstF_maskf_inv, dstB_maskb_inv, dstF0_maskb, dstB0_maskf;
if (sizeof(PixelType) == 1) {
dstF_maskf_inv = _mm256_mullo_epi16(dstF, maskf_inv);
dstB_maskb_inv = _mm256_mullo_epi16(dstB, maskb_inv);
dstF0_maskb = _mm256_mullo_epi16(dstF0, maskb);
dstB0_maskf = _mm256_mullo_epi16(dstB0, maskf);
} else {
dstF_maskf_inv = _mm256_mullo_epi32(dstF, maskf_inv);
dstB_maskb_inv = _mm256_mullo_epi32(dstB, maskb_inv);
dstF0_maskb = _mm256_mullo_epi32(dstF0, maskb);
dstB0_maskf = _mm256_mullo_epi32(dstB0, maskf);
}
__m256i f = _mm256_add_epi32(dstF0_maskb, dstB_maskb_inv);
__m256i b = _mm256_add_epi32(dstB0_maskf, dstF_maskf_inv);
if (sizeof(PixelType) == 1) {
f = _mm256_mullo_epi32(f, maskf);
b = _mm256_mullo_epi32(b, maskb);
f = _mm256_add_epi32(f, dwords_255);
b = _mm256_add_epi32(b, dwords_255);
f = _mm256_srai_epi32(f, 8);
b = _mm256_srai_epi32(b, 8);
} else {
const __m256i qwords_255 = _mm256_set1_epi64x(255);
__m256i tempf = _mm256_mul_epu32(f, maskf);
__m256i tempb = _mm256_mul_epu32(b, maskb);
tempf = _mm256_add_epi64(tempf, qwords_255);
tempb = _mm256_add_epi64(tempb, qwords_255);
tempf = _mm256_srli_epi64(tempf, 8);
tempb = _mm256_srli_epi64(tempb, 8);
f = _mm256_srli_epi64(f, 32);
b = _mm256_srli_epi64(b, 32);
f = _mm256_mul_epu32(f, _mm256_srli_epi64(maskf, 32));
b = _mm256_mul_epu32(b, _mm256_srli_epi64(maskb, 32));
f = _mm256_add_epi64(f, qwords_255);
b = _mm256_add_epi64(b, qwords_255);
f = _mm256_srli_epi64(f, 8);
b = _mm256_srli_epi64(b, 8);
f = _mm256_or_si256(tempf, _mm256_slli_epi64(f, 32));
b = _mm256_or_si256(tempb, _mm256_slli_epi64(b, 32));
}
f = _mm256_add_epi32(f, dstF_maskf_inv);
b = _mm256_add_epi32(b, dstB_maskb_inv);
f = _mm256_add_epi32(f, dwords_255);
b = _mm256_add_epi32(b, dwords_255);
f = _mm256_srai_epi32(f, 8);
b = _mm256_srai_epi32(b, 8);
if (sizeof(PixelType) == 1) {
f = _mm256_madd_epi16(f, dwords_256_time256);
b = _mm256_madd_epi16(b, dwords_time256);
} else {
f = _mm256_mullo_epi32(f, dwords_256_time256);
b = _mm256_mullo_epi32(b, dwords_time256);
}
__m256i dst = _mm256_add_epi32(f, b);
dst = _mm256_srai_epi32(dst, 8);
dst = _mm256_packus_epi32(dst, dst);
dst = _mm256_permute4x64_epi64(dst, 0xe8); // 0b11101000 - copy third qword to second qword
__m128i dst128 = _mm256_castsi256_si128(dst);
if (sizeof(PixelType) == 1) {
dst128 = _mm_packus_epi16(dst128, dst128);
_mm_storel_epi64((__m128i *)&pdst[w], dst128);
} else {
_mm_storeu_si128((__m128i *)&pdst[w], dst128);
}
}
int64_t * const restrict del_pr = _del_pr+PAD;
#ifdef PARASAIL_TABLE
parasail_result_t *result = parasail_result_new_table1(s1Len, s2Len);
#else
#ifdef PARASAIL_ROWCOL
parasail_result_t *result = parasail_result_new_rowcol1(s1Len, s2Len);
#else
parasail_result_t *result = parasail_result_new();
#endif
#endif
int32_t i = 0;
int32_t j = 0;
int32_t end_query = 0;
int32_t end_ref = 0;
int64_t score = NEG_INF;
__m256i vNegInf = _mm256_set1_epi64x(NEG_INF);
__m256i vNegInf0 = _mm256_srli_si256_rpl(vNegInf, 8); /* shift in a 0 */
__m256i vOpen = _mm256_set1_epi64x(open);
__m256i vGap = _mm256_set1_epi64x(gap);
__m256i vOne = _mm256_set1_epi64x(1);
__m256i vN = _mm256_set1_epi64x(N);
__m256i vNegOne = _mm256_set1_epi64x(-1);
__m256i vI = _mm256_set_epi64x(0,1,2,3);
__m256i vJreset = _mm256_set_epi64x(0,-1,-2,-3);
__m256i vMaxScore = vNegInf;
__m256i vEndI = vNegInf;
__m256i vEndJ = vNegInf;
__m256i vILimit = _mm256_set1_epi64x(s1Len);
__m256i vILimit1 = _mm256_sub_epi64(vILimit, vOne);
__m256i vJLimit = _mm256_set1_epi64x(s2Len);
__m256i vJLimit1 = _mm256_sub_epi64(vJLimit, vOne);
void Initialize() {
/*
Round constants for p_1:
01, 02, 05, 0a, 15, 0b, 17, 0e, 1d, 1b, 16, 0c
Round constants for p_2:
18, 11, 03, 07, 0f, 1f
Round constants for p_3:
1e, 1c, 19, 13, 06, 0d
*/
shuffleControlMaskFirstReg = _mm256_setr_epi8(
0, 1, 2, 3, 4, 5, 6, 7, //0
9, 10, 11, 12, 13, 14, 15, 8, //1
18, 19, 20, 21, 22, 23, 16, 17, //2
28, 29, 30, 31, 24, 25, 26, 27); //4
shuffleControlMaskSecondReg = _mm256_setr_epi8(
7, 0, 1, 2, 3, 4, 5, 6, //7
255, 255, 255, 255, 255, 255, 255, 255, //Setting it to 0xFF makes shuffle zero the bits
255, 255, 255, 255, 255, 255, 255, 255, //Setting it to 0xFF makes shuffle zero the bits
255, 255, 255, 255, 255, 255, 255, 255); //Setting it to 0xFF makes shuffle zero the bits
invShuffleControlMaskFirstReg = _mm256_setr_epi8(
0, 1, 2, 3, 4, 5, 6, 7, //0
15, 8, 9, 10, 11, 12, 13, 14, //1
22, 23, 16, 17, 18, 19, 20, 21, //2
28, 29, 30, 31, 24, 25, 26, 27); //4
invShuffleControlMaskSecondReg = _mm256_setr_epi8(
1, 2, 3, 4, 5, 6, 7, 0, //7
255, 255, 255, 255, 255, 255, 255, 255, //Setting it to 0xFF makes shuffle zero the bits
255, 255, 255, 255, 255, 255, 255, 255,
255, 255, 255, 255, 255, 255, 255, 255);
m256iAllOne = _mm256_set1_epi64x(0xFFFFFFFFFFFFFFFF);
//Set the bits to 1111'1111 in the column two, second row byte, if the roundconstant has a onebit on this indice
//p1
p1_constants_bit0[0] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[1] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit0[2] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[3] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit0[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[7] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit0[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit0[10] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit0[11] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit1[0] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit1[1] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[2] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit1[3] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[4] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit1[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[7] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[8] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit1[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[10] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit1[11] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit2[0] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit2[1] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit2[2] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit2[3] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit2[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit2[5] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit2[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit2[7] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit2[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit2[9] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit2[10] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit2[11] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit3[0] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit3[1] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit3[2] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit3[3] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit3[4] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit3[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit3[6] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit3[7] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit3[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit3[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit3[10] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit3[11] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit4[0] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit4[1] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit4[2] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit4[3] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit4[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit4[5] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit4[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit4[7] = _mm256_set_epi64x(0, 0, 0, 0);
p1_constants_bit4[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit4[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit4[10] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p1_constants_bit4[11] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit0[0] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit0[1] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[2] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[3] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[6] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit0[7] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit0[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit0[10] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit0[11] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[0] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit1[1] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit1[2] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[3] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[7] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit1[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[9] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit1[10] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit1[11] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[0] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit2[1] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit2[2] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit2[3] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[7] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[8] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit2[9] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit2[10] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit2[11] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[0] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[1] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit3[2] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit3[3] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit3[4] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[7] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[8] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit3[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit3[10] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit3[11] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[0] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[1] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[2] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit4[3] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit4[4] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit4[5] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[6] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[7] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[8] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[9] = _mm256_set_epi64x(0, 0, 0b0000000000000000000000000000000000000000000000001111111100000000, 0);
p4_constants_bit4[10] = _mm256_set_epi64x(0, 0, 0, 0);
p4_constants_bit4[11] = _mm256_set_epi64x(0, 0, 0, 0);
}